Holmes J.B., Waters H.S., & Rajaram S. (1998). The phenomenology of false memories: Episodic content and confidence. Jounal of Experimental Psychology: Learning, Memory, and Cognition, 24, 95-107.

• Subjects heard a list of sentences that were similar in idea units but differed in levels of complexity. They were asked a semantic orienting question directly after each sentence. Later, a recognition task was performed followed by confidence ratings for each item recognized.
• The semantic integration paradigm infers that when two similar sentences are encountered, they are semantically combined in memory to form a gist integration of the sentences. When subjects were asked if they heard a similar non-presented sentence they confidently agreed. This rating of higher confidence despite presentation or non-presentation is termed the linear abstraction effect.
• Results: Subjects couldn’t distinguish the old sentences from the new ones. Linear abstraction effects occurred as the levels of sentence complexity increased, most likely because the more detailed sentences closely matched the subjects’ gist integrations.

• Using Remember-Know judgments, they found that people are very confident in their false alarms and that these false alarms are rated as Remember judgments.
• Remember judgments must be made due to episodic content of the memory. Is that possible if the event never even occurred? This leads to the question of whether false memories can have similar phenomenological characteristics of true memories, and thus, the following studies.

* Goal: To replicate the Bransford and Franks (1971) false memory effect caused by integrative processes and to assess recognition using Remember-Know judgments.

* Remember judgments are used to measure episodic content, the feeling that an event has been experienced in the past.

*Prediction: False memories that arise from integrative processes will be given Remember judgments indicating the existence of episodic content.

 Method

*Four sentences were taken from the Bransford & Franks (1971) study. These were used to form four acquisition lists (24 sentences) and four recognition lists (48 sentences--half occurred and half did not). A total of 12 sentences contained similar idea units for each of the original sentences.

*After each sentence in the acquisition phase, a semantic orienting question was asked.

*Following the acquisition phase, filler tasks resulted in a 20 min. retention interval.

*Recognition test included Remember-Know judgments.

 Results

*Replicated Bransford and Franks’s results for linear abstraction effects and poor discrimination between old and new sentences.

*The same effects occurred for hits and false alarms, but no effects for Know judgments.

 Discussion

1) Subjects were more likely to give a Remember rating if the sentence ideas had integrated strongly. Integration must have occurred during Know judgments since no discrimination was made b/w old and new sentences. This suggests that memories with linear abstraction effects must have episodic content.

2) The large number of Remember false alarms indicate that people consciously knew they encountered non-presented sentences and had associated memories. Thus, false memories were more likely to have episodic content due to semantic integration processes.

 Proposal

Semantic integration must occur by a two-step process: (1) Semantic integration processes result in an integrated gist representation and transference of the episodic content from the sentence to the gist representation and (2) The episodic content increases the person’s confidence in their memories.

*Goal: To replicate Exp. 1 results and to explore the role of episodic content in confidence judgments combining the Remember-Know distinction and Bransford and Franks’s paradigm.

*Prediction: Subjects will have higher confidence with both Remember judgments and with greater sentence complexity.

 Method

The materials are the same as in exp. 1, but include a 5-point confidence scale.

 Results

1) The results of Exp. 1 were replicated proving its reliability.

2) The effects of confidence ratings on performance were not adverse, so these ratings could be used in future studies.

 Discussion

Even if episodic content didn’t exit, subjects may have matched their gist representations to the more complex sentences causing greater linear abstraction effects for confidence ratings.

*Goal: Need to reduce levels of semantic integration and show that confidence relies only on episodic content.

*Hypothesis: Confidence is reliant on both the processes of episodic content plus matching recognition items and the gist representation.

Same as Exp. 2 with some changes: Instead of having them answer a question after each sentence, they counted the number of letters in the last word of the sentence and wrote the number on the answer sheet.

* Changing encoding conditions caused a reduction of semantic integration effects, but didn’t eliminate semantic processing. The number of Remember hits and false alarms was reduced indicating that altered encoding processes led to a reduction in semantic integration. (Fig. 5 and Fig. 8 show the differences).

* Linear abstraction effects on confidence ratings were eliminated, and confidence for Remember items was greater than confidence for Know items across all levels of complexity (flattening effect)

• Four Principle Results:

2) Most memories arising from strong integrative processes contained episodic content.

3) A large amount of Remember false alarms indicated that people had conscious awareness of encountering a non-presented item with supporting memories.

4) Ratings of confidence rely on both the presence of episodic content and the degree of match b/w recognition items and gist representations.

• Why do false memories have episodic content?

* Memory can be a constructive process, where retrieved memories are reconstructions rather than exact reproductions. This explains why semantic integration may occur, resulting in a gist representation.

* It is also possible that a memory can occur from first-person or third-person point of view which causes different representations to arise. Therefore the reconstructive nature of episodic memory may be the cause of inaccurate binding and false memories.

• Reality monitoring theory

1) If a memory is perceived to be accurate, it should have the same phenomenonological characteristics of the actual event.

2) Subjects may fail to attend to reality monitoring cues, which indicate whether an event may have actually occurred. This is not likely because if a Remember judgment is made, such cues would have to be noted and somehow connected to false memories.

3) False memories should lack these cues if they never existed in the first place. This is not the case. Paying close attention to cues may cause a person to attribute them to a false memory and say that it did occur.

• The accuracy of memory cannot be assessed based on both episodic content and high levels of confidence that a memory is true.
• The authors support the idea that inaccurate binding of episodic content to false memories may occur. So, good reality monitors should judge a false memory to be true.
• Further study is needed to investigate these assumptions.